From mboxrd@z Thu Jan 1 00:00:00 1970 From: Richard Smith Subject: 2.6.x scsi software raid IO tuning Date: Wed, 29 Oct 2003 16:00:02 -0600 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <3FA03862.5080203@bitworks.com> Reply-To: rsmith@bitworks.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from b44.xdsl.pgtc.com ([198.70.248.44]:46841 "HELO bitworks.com") by vger.kernel.org with SMTP id S261699AbTJ2WAU (ORCPT ); Wed, 29 Oct 2003 17:00:20 -0500 Received: from bitworks.com [192.168.1.22] by bitworks.com [127.0.0.1] with SMTP (MDaemon.v2.7.SP4.R) for ; Wed, 29 Oct 2003 15:58:59 -0600 List-Id: linux-scsi@vger.kernel.org To: linux-scsi@vger.kernel.org I receintly came across several SCSI racks with drives. So I've been playing with various software RAID setups with between 8 - 15 drives. I have a sym8xx SE chain and a AIC7xxx HVD chain. Using Bonnie++ to measure IO throughput of my RAID 5 setup. I also decided to try the latest 2.6 test kernels. I built the 2.6 with the same setup as my 2.4.22 kernel. Preemption was disabled and I believe that I have all debugging output turned off. I found that in every test case I ran Bonnie++ reported the 2.6 kernel IO as 3 to 5 MB/s slower than the 2.4 kernel. File creation/deletion was also slower. Is there some tuning knobs that I need to tweak or some option that might be enabled causing this? I thought the 2.6 was supposed to have much better IO throughput when setup in non-preempt. So I expected to at least get the same numbers not less. My apologies if this is a md issue rather than a scsi issue but out of my info searches so far this list seemed like the best place to ask for info. -- Richard A. Smith rsmith@bitworks.com